Gesture recognition is becoming a more common interaction tool in the fields of ubiquitous and wearable computing. Designing a system to perform gesture recognition, however, can be a cumbersome task. Hidden Markov models (HMMs), a pattern recognition technique commonly used in speech recognition, can be used for recognizing certain classes of gestures. Existing HMM toolkits for speech recognition can be adapted to perform gesture recognition, but doing so requires significant knowledge of the speech recognition literature and its relation to gesture recognition. This paper introduces the Georgia Tech Gesture Toolkit GT 2 k which leverages Cambridge University’s speech recognition toolkit, HTK, to provide tools that support gesture recognition research. GT 2 k provides capabilities for training models and allows for both real–time and off-line recognition. This paper presents four ongoing projects that utilize the toolkit in a variety of domains.

We introduce a new dynamic model with the capability of recognizing both activities that an individual is performing as well as where that individual is located. Our approach is novel in that it utilizes a dynamic graphical model to jointly estimate both activity and spatial context over time based on the simultaneous use of asynchronous observations consisting of GPS measurements, and a small mountable sensor board. Joint inference is quite desirable as it has the ability to improve accuracy of the model and consistency of the location and activity estimates. The parameters of our model are trained on partially labeled data. We apply virtual evidence to improve data annotation, giving the user high flexibility when labeling training data. We present results indicating the performance gains achieved by virtual evidence for data annotation and the joint inference performed by our system. 1

In this paper, we present a real-time algorithm for automatic recognition of not only physical activities, but also, in some cases, their intensities, using five triaxial wireless accelerometers and a wireless heart rate monitor. The algorithm has been evaluated using datasets consisting of 30 physical gymnasium activities collected from a total of 21 people at two different labs. On these activities, we have obtained a recognition accuracy performance of 94.6 % using subject-dependent training and 56.3 % using subjectindependent training. The addition of heart rate data improves subject-dependent recognition accuracy only by 1.2 % and subject-independent recognition only by 2.1%. When recognizing activity type without differentiating intensity levels, we obtain a subjectindependent performance of 80.6%. We discuss why heart rate data has such little discriminatory power. 1.

www.wearable.ethz.ch Abstract. The paper describes a method that allows us to derive the location of an acceleration sensor placed on the user’s body solely based on the sensor’s signal. The approach described here constitutes a first step in our work towards the use of sensors integrated in standard appliances and accessories carried by the user for complex context recognition. It is also motivated by the fact that device location is an important context (e.g. glasses being worn vs. glasses in a jacket pocket). Our method uses a (sensor) location and orientation invariant algorithm to identify time periods where the user is walking and then leverages the specific characteristics of walking motion to determine the location of the body-worn sensor. In the paper we outline the relevance of sensor location recognition for appliance based context awareness and then describe the details of the method. Finally, we present the results of an experimental study with six subjects and 90 walking sections spread over several hours indicating that reliable recognition is feasible. The results are in the low nineties for frame by frame recognition and reach 100 % for the more relevant event based case. 1

The paper deals with the design of a sound recognition system focused on an ultra low power hardware implementation in a button like miniature form factor. We present the results of the first design phase focused on selection and experimental evaluation of sound classes and algorithms suitable for low power realization. We also present the VHDL model of the hardware showing that our method can be implemented with minimal resources. Our approach is based on spectrum analysis to distinguish between a subset of sound sources with a clear audio signature. It also uses intensity analysis from microphones placed at different locations to correlate the sounds with user activity. 1.

In recent years much work has been done on human activity recognition using wearable sensors. As we begin to deploy hundreds and even thousands of wearable sensors on regular workers, hospital patients, and army soldiers, the question shifts more toward a balance between what information can be gained and their broad immediate user acceptance. In this paper we compare the activity classification accuracy of four different configurations of accelerometer placement on the human body using hidden Markov models (HMMs). We find the classification accuracy of a single accelerometer placed in three different parts of the body and evaluate whether there is a significant improvement in recognition accuracy by adding multiple accelerometers or not. We also find the number of hidden states that best models each activity by achieving the lowest test error using K-fold cross-validation.

With the availability of affordable sensors and sensor networks, sensor-based human activity recognition has attracted much attention in artificial intelligence and ubiquitous computing. In this paper, we present a novel two-phase approach for detecting abnormal activities based on wireless sensors attached to a human body. Detecting abnormal activities is a particular important task in security monitoring and healthcare applications of sensor networks, among many others. Traditional approaches to this problem suffer from a high false positive rate, particularly when the collected sensor data are biased towards normal data while the abnormal events are rare. Therefore, there is a lack of training data for many traditional data mining methods to be applied. To solve this problem, our approach first employs a one-class support vector machine (SVM) that is trained on commonly available normal activities, which filters out the activities that have a very high probability of being normal. We then derive abnormal activity models from a general normal model via a kernel nonlinear regression (KNLR) to reduce false positive rate in an unsupervised manner. We show that our approach provides a good tradeoff between abnormality detection rate and false alarm rate, and allows abnormal activity models to be automatically derived without the need to explicitly label the abnormal training data, which are scarce. We demonstrate

Recent advances in wearable sensing and computing devices and in fast probabilistic inference techniques make possible the fine-grained estimation of a person’s activities over extended periods of time [6]. Such technologies enable applications ranging from context aware computing to support for cognitively

Wearable computing aims to empower people by providing intelligence embedded within garments. It relies on sensors placed at different locations of the body. To foster useracceptance sensors should be small, light, and unobtrusive. In this paper we present a wearable platform that addresses those challenges: a miniature networked SensorButton with the form factor of a button, so that it can be integrated in garments in an unobtrusive way. It has several sensors used in wearable computing, on-board processing power, a wireless link for sensor networking or communication with a base station, and it focuses on low power consumption. We describe its use to recognize user activity and highlight the need for further research in poweraware algorithms for wearable computing.

We propose a distributed recognition framework to classify continuous human actions using a low-bandwidth wearable motion sensor network, called distributed sparsity classifier (DSC). The algorithm classifies human actions using a set of training motion sequences as prior examples. It is also capable of rejecting outlying actions that are not in the training categories. The classification is operated in a distributed fashion on individual sensor nodes and a base station computer. We model the distribution of multiple action classes as a mixture subspace model, one subspace for each action class. Given a new test sample, we seek the sparsest linear representation of the sample w.r.t. all training examples. We show that the dominant coefficients in the representation only correspond to the action class of the test sample, and hence its membership is encoded in the sparse representation. Fast linear solvers are provided to compute such representation via ℓ 1-minimization. To validate the accuracy of the framework, a public wearable action recognition database is constructed, called wearable action recognition database (WARD). The database is comprised of 20 human subjects in 13 action categories. Using up to five motion sensors in the WARD database, DSC achieves state-of-the-art performance. We further show that the recognition precision only decreases gracefully using smaller subsets of active sensors. It validates the robustness of the distributed recognition framework on an unreliable wireless network. It also demonstrates the ability of DSC to conserve sensor energy for communication while preserve accurate global classification.